editorial-review - SKILL.md Agent Skill

name: editorial-review description: Review a document for clarity, consistency, correctness, and tone

Editorial Review

When to Use

Use this skill when tasked with reviewing documentation, reports, or any written deliverable for quality before it is shared or published.

Procedure

1. First Pass -- Read Without Changes

Read the entire document without making any changes. Note your overall impression:

DOC_FILE="${1:-/home/shared/document.md}"

# Read the full document
cat "$DOC_FILE"

# Get basic stats
echo ""
echo "=== Document Stats ==="
wc -l "$DOC_FILE" | awk '{print "Lines:", $1}'
wc -w "$DOC_FILE" | awk '{print "Words:", $1}'
grep -c '^#' "$DOC_FILE" | xargs -I{} echo "Headings: {}"
grep -c '```' "$DOC_FILE" | awk '{print "Code blocks:", $1/2}'
grep -cE '^\|' "$DOC_FILE" | xargs -I{} echo "Table rows: {}"

After reading, answer these questions before proceeding:

What is the document's purpose?
Who is the intended audience?
Does the structure make sense for that purpose?

2. Second Pass -- Detailed Review

Check each of these dimensions:

Clarity

Can each sentence be understood on first reading?
Are technical terms defined on first use?
Are there ambiguous pronouns ("it", "this", "that") without clear antecedents?
Are instructions specific enough to follow without guessing?

# Find potentially ambiguous sentences (very long sentences)
awk 'length > 120 && !/^[|#`]/' "$DOC_FILE" | head -10
echo "(Sentences over 120 chars may need splitting)"

# Find undefined acronyms
grep -oE '\b[A-Z]{2,6}\b' "$DOC_FILE" | sort -u | while read acr; do
  FIRST=$(grep -n "\b$acr\b" "$DOC_FILE" | head -1 | cut -d: -f1)
  DEFINED=$(grep -n "$acr (" "$DOC_FILE" | head -1 | cut -d: -f1)
  if [ -z "$DEFINED" ] && [ "$(echo "$acr" | wc -c)" -gt 3 ]; then
    echo "Acronym $acr first used on line $FIRST — not defined"
  fi
done

Consistency

Are terms used consistently (not "config" in one place and "configuration" in another)?
Are formatting conventions consistent (heading levels, code block style, list style)?
Are command examples consistent with the project's actual commands?

# Check for inconsistent terminology
for pair in "config:configuration" "dir:directory" "repo:repository" "env:environment" "arg:argument"; do
  SHORT=$(echo "$pair" | cut -d: -f1)
  LONG=$(echo "$pair" | cut -d: -f2)
  S_COUNT=$(grep -ci "\b${SHORT}\b" "$DOC_FILE" 2>/dev/null)
  L_COUNT=$(grep -ci "\b${LONG}\b" "$DOC_FILE" 2>/dev/null)
  if [ "$S_COUNT" -gt 0 ] && [ "$L_COUNT" -gt 0 ]; then
    echo "Inconsistent: '$SHORT' ($S_COUNT) vs '$LONG' ($L_COUNT) — pick one"
  fi
done

# Check heading level consistency
grep '^#' "$DOC_FILE" | awk '{print length($1), $0}' | head -20
echo "(Verify heading levels are sequential — no jumps from ## to ####)"

Correctness

Are facts accurate?
Do code examples work?
Are file paths, command names, and flag names correct?
Do cross-references point to things that exist?

# Verify referenced file paths exist
grep -oE '`[~/][a-zA-Z0-9_./-]+`' "$DOC_FILE" | tr -d '`' | while read path; do
  expanded=$(eval echo "$path" 2>/dev/null)
  [ -e "$expanded" ] && echo "OK: $path" || echo "MISSING: $path"
done

# Verify referenced commands exist
grep -oE '`(bash|node|python3|npm|jq|curl) [^`]+`' "$DOC_FILE" | tr -d '`' | while read cmd; do
  BINARY=$(echo "$cmd" | awk '{print $1}')
  which "$BINARY" >/dev/null 2>&1 && echo "OK: $BINARY" || echo "MISSING: $BINARY"
done

Completeness

Are there gaps where a reader would be stuck?
Does every "how" have a "why"?
Are error cases and edge cases documented?
Is there a troubleshooting section for likely problems?

Tone

Is the tone appropriate for the audience?
Is it consistent throughout (not formal in intro and casual in body)?
Does it avoid condescension ("simply", "just", "obviously")?

# Check for condescending language
grep -niE '\b(simply|just|obviously|clearly|trivially|of course|easy|easily)\b' "$DOC_FILE" \
  | head -10
echo "(Consider removing these — they imply the reader should find it obvious)"

3. Produce Review Notes

For each issue found, format as:

### Issue [N]: [Category]
- **Location:** Line [N] / Section "[heading]"
- **Current:** "[quoted text as it appears]"
- **Suggested:** "[replacement text]"
- **Reason:** [why this change improves the document]

4. Write the Review

REVIEW_FILE="/home/shared/review-$(date +%Y%m%d)-$(basename "$DOC_FILE" .md).md"

cat > "$REVIEW_FILE" <<'EOF'
# Editorial Review

**Document:** [document name]
**Reviewer:** [agent name]
**Date:** YYYY-MM-DD

## Overall Assessment
[2-3 sentences: is this document ready to publish? What is its biggest strength? What needs the most work?]

## Summary
| Category | Issues Found |
|----------|-------------|
| Clarity | N |
| Consistency | N |
| Correctness | N |
| Completeness | N |
| Tone | N |

## Issues

### Issue 1: [Category]
- **Location:** ...
- **Current:** "..."
- **Suggested:** "..."
- **Reason:** ...

## Positive Notes
- [Something the document does well — preserve the author's voice]
- [Another strength]
EOF

# Register as artifact
bash /home/shared/scripts/artifact.sh register \
  --name "review-$(basename "$DOC_FILE" .md)" \
  --type "review" \
  --path "$REVIEW_FILE" \
  --description "Editorial review of $(basename "$DOC_FILE")"

Quality Checklist

Full document read before any edits were made
All five dimensions checked (clarity, consistency, correctness, completeness, tone)
Every issue has location, current text, suggested text, and reason
Code examples verified to be runnable
File paths and command references verified to exist
Positive aspects acknowledged (not just criticism)
Author's voice and style preserved in suggestions
Review is actionable (author can apply each suggestion without guessing)